GitFlic Code Dataset

A comprehensive code dataset compiled from GitFlic, Russia's first native service for storing and working with source code. This dataset is specifically designed to support training code models with strong Russian language understanding and authentic Russian coding practices.

---

Overview

The GitFlic Code Dataset represents a significant code corpus from Russia's leading Git-based code hosting platform, capturing diverse projects across 690 programming languages. It serves as a valuable resource for developing multilingual code understanding models reflecting Russian development practices and standards.

Key Statistics

Metric	Value
Total Files	5,975,978
Total Repositories	12,527
Compressed Size	6.44 GB (Parquet with Zstd)
Programming Languages	690
File Format	6 Parquet files

---

Dataset Characteristics

Scope and Coverage

This dataset captures code from over 12,500 repositories hosted on GitFlic, including:

• Russian code community: Extensive coverage of code written by Russian developers and enterprises, featuring Russian comments, documentation, and variable naming conventions
• Diverse language ecosystem: Support for 690 distinct programming languages
• Developer and enterprise projects: A comprehensive mix of individual developer projects and enterprise-grade codebases
• Quality-assured: Deduplicated and filtered to remove binary files and non-code content

Programming Languages

The dataset encompasses 690 languages. The 30 most represented languages by file count are:

Rank	Language	File Count
1	C	739,012
2	Java	634,899
3	C++	587,528
4	JavaScript	422,832
5	PHP	365,105
6	XML	291,920
7	Markdown	211,574
8	Shell	207,178
9	Python	206,443
10	Unity3D Asset	150,654
11	SVG	150,136
12	TypeScript	141,886
13	Text	139,406
14	JSON	126,214
15	HTML	122,341
16	Go	109,740
17	YAML	89,416
18	Roff	82,609
19	C#	77,520
20	Makefile	63,594
21	LLVM	55,680
22	Scala	53,395
23	Unix Assembly	49,909
24	Rust	35,553
25	reStructuredText	35,023
26	Objective-C	34,151
27	Ruby	33,366
28	CMake	33,030
29	CSS	31,664
30	TSX	31,397

---

Dataset Structure

Data Fields

Each record contains three fields providing content and metadata:

Field	Type	Description
file_text	string	Complete file content in UTF-8 encoding
language	string	Programming language identified using github-linguist
file_name	string	Unique file identifier within the dataset

Sample Record

{
    "file_text": "package com.example.demo;\n\nimport org.springframework.boot.SpringApplication;\n...",
    "language": "Java",
    "file_name": "Application.java"
}

File Format

• Format: Apache Parquet with Zstd compression
• Structure: 6 consolidated files (gitflic-00000.parquet to gitflic-00005.parquet)
• Encoding: UTF-8
• Split: All examples are included in a single training split (no validation or test splits)

---

Data Creation Process

Language Detection Methodology

Programming languages are identified using github-linguist, GitHub's robust library for language detection. This ensures consistent and reliable classification across all files in the dataset.

Source Data

All data originates from public repositories hosted on GitFlic, Russia's native code hosting platform.

Quality Filtering

The dataset has undergone systematic filtering to ensure quality and usability:

Deduplication

• Files have been deduplicated to ensure each code file appears only once in the dataset

Binary File Removal

• Binary files have been systematically excluded from the dataset
• Only text-based source code files are retained

Content Quality

• Files filtered to ensure data quality and remove non-code content
• UTF-8 encoding validation applied to all text files

---

Usage Considerations

Data Privacy and Security

The dataset may contain sensitive information that requires careful handling:

• Email Addresses: Present in code comments, documentation, or configuration files
• Credentials: Accidentally committed API keys or authentication tokens
• Personal Information: Names, phone numbers, and other identifiable data in comments or documentation

Users should implement appropriate filtering and anonymization when preparing data for model training.

Licensing and Attribution

This dataset aggregates source code from repositories with diverse licenses. Any use of code or data derived from this dataset must comply with the original repository licenses, including attribution requirements where applicable.

Users are responsible for:

• Reviewing applicable license terms for code they utilize
• Providing proper attribution when required
• Ensuring compliance with license restrictions
• Respecting the rights of original authors and using the data responsibly

---

Technical Details

Source: Public repositories hosted on GitFlic

Annotations: Machine-generated (language detection)

Multilingual Support: Includes multilingual code and documentation with emphasis on Russian content

Task Categories: Text generation, code modeling, language understanding

Tags: Code, Russian language, multilingual, open-source development

---